OHDSI Center Events
CBER BEST Seminar Series: Bayesian Adaptive Validation Design and Extensions to Vaccine Surveillance (POSTPONED)
11:00 am – 3:00 pm
This event has been postponed. We will post the new date once it has been confirmed.
Anyone can register and join for free.
The CBER BEST Initiative Seminar Series is designed to share and discuss recent research of relevance to ongoing and future surveillance activities of CBER-regulated products, namely biologics. The seminars will provide information on characteristics of biologics, required infrastructure, study designs, and analytic methods utilized for pharmacovigilance and pharmacoepidemiologic studies of biologics. They will also cover information regarding potential data sources, informatics challenges and requirements, utilization of real-world data and evidence, and risk-benefit analysis for biologic products. The length of each session may vary, and the presenters will be invited from outside FDA.
Bayesian adaptive validation design and extensions to vaccine surveillance
Timothy Lash (Emory University)
Lindsay Collin (University of Utah)
An internal validation substudy compares an imperfect measurement of a variable with a gold-standard measurement in a subset of the study population. Existing guidance on optimal sampling for validation substudies has assumed complete enrollment and follow-up for the entire cohort, followed by collection of validation data for a sample of fixed size from the complete population. The adaptive validation approach provides a framework to monitor validation data as they accrue until predefined stopping criteria for the goals of validation are met. The method provides an approach to effective and efficient estimation of classification parameters as validation data accrue, sometimes synchronously with enrollment and primary data collection for the entire cohort. The adaptive validation design has great potential for studies of vaccine or drug safety that use real world data to ascertain adverse events, sometimes recorded with error in these data.